training_args.json

{
  "model": "phi4-mini",
  "use_unsloth_model": true,
  "dataset": "UWV/wim_instruct_signaalberichten_to_jsonld_agent_steps",
  "max_samples": null,
  "filter_n5_only": true,
  "max_seq_length": 4096,
  "batch_size": 8,
  "gradient_accumulation_steps": 1,
  "max_steps": 1735,
  "learning_rate": 2e-05,
  "warmup_steps": 50,
  "max_grad_norm": 1.0,
  "lora_r": 512,
  "lora_alpha": 1024,
  "lora_dropout": 0.1,
  "target_modules": [
    "q_proj",
    "k_proj",
    "v_proj",
    "o_proj"
  ],
  "load_in_4bit": true,
  "use_gradient_checkpointing": false,
  "rope_scaling": null,
  "output_dir": "/data/model_outputs/n5_label_addition_model",
  "assistant_only_loss": true,
  "no_eval": false,
  "num_workers": 8,
  "full_finetune": false,
  "wandb": true,
  "seed": 42,
  "push_to_hub": false,
  "hub_model_id": null
}